Studying parallel evolutionary algorithms: The cellular programming case

Abstract

Parallel evolutionary algorithms, studied to some extent over the past few years, have proven empirically worthwhile-though there seems to be lacking a better understanding of their workings. In this paper we concentrate on cellular (fine-grained) models, presenting a number of statistical measures, both at the genotypic and phenotypic levels. We demonstrate the application and utility of these measures on a specific example, that of the cellular programming evolutionary algorithm, when used to evolve solutions to a hard problem in the cellular-automata domain, known as synchronization.